Re: SUO: a silly question about the new modular architecture
Leo,
Saying there is no alternative is just too strong. There are valid
arguments for both a single ontology and a multiple ontology
approach. Reconciling different theories might be a matter of creating
bridges or translation to legacy ontologies or systems just as might be
done from a data warehouse to legacy database.
Adam
At 09:05 PM 9/7/2001 -0400, Leo Obrst wrote:
>John S. is right. There really is no alternative to the modular
>approach. We need to make that work. Do we need to do this immediately?
>No. But we need to build that approach into our methodology and intent
>soon. Again, there is really no alternative, since the semantics one
>needs will always vary. Eventually one theory or another (perdurantist
>or mereotopological #3) may win out, but we will always need to
>reconcile (i.e., represent and use) different theories.
>
>Leo
>
>"John F. Sowa" wrote:
> >
> > John and Adam,
> >
> > The modular approach would minimize the amount of overlap, so there
> > would actually be very little recoding that would need to be done.
> > For example, all the axioms about geometry, space, time, etc.,
> > could be used unchanged. Axioms for mathematical structures,
> > physics, engineering, etc., would be unchanged if they used the
> > 4D coordinate system.
> >
> > AP > I'm inclined to think that
> > > there will be a big ripple effect which will mean we'll have two
> > > ontologies.
> >
> > That is just one more reason why it is important to have a modular
> > approach. It enables you to determine exactly where the dependencies
> > are for any particular subtheory.
> >
> > AP> The more divisions we have, the less of a standard we
> > > have.
> >
> > That all depends on how you organize your theor(y/ies). With the
> > lattice, there is a place for everything, and the dependencies
> > are shown by the partial ordering (which shows the explicit
> > inheritance paths from one theory to another). The separate
> > modules have the effect of firewalls that isolate or restrain
> > the ripple effects to just those subtheories that inherit from
> > the one that was changed.
> >
> > JT>Is the theory that one could pick either the 3d or 4d spatio-temporal
> > >module and then use the rest of the ontology modules as is (they will work
> > >equally well with either the 3d or 4d module), or would all the other
> > >modules in the system have to be encoded twice -- one version to work with
> > >the 3d module, and another version to work with the 4d module?
> >
> > With a single monolithic ontology, a change to one axiom creates
> > a totally new ontology, which requires everything to be tested
> > to ensure consistency. You can't do unit testing on individual
> > modules when the entire ontology is a single monolith.
> >
> > If the modules are organized in a partial ordering (similar to
> > the modules of an object-oriented programming language), then
> > the only modules that are affected by a change to theory X are
> > the ones that inherit from X.
> >
> > Bottom line: The lattice of theories supports multiple inheritance
> > in the same way as O-O inheritance. With the monolithic approach,
> > every change creates a new monolith.
> >
> > John Sowa
>
>--
>_____________________________________________
>Dr. Leo Obrst The MITRE Corporation
>mailto:lobrst@mitre.org Intelligent Information Management/Exploitation
>Voice: 703-883-6770 7515 Colshire Drive, M/S W640
>Fax: 703-883-1379 McLean, VA 22102-7508, USA
Adam Pease
Teknowledge
(650) 424-0500 x571